1. Technical Field
The present invention relates to a technology of producing three-dimensional model data, which is used in processing of acquiring three-dimensional information from image data obtained by imaging an object with plural cameras and checking the acquired three-dimensional information against the three-dimensional model data to perform three-dimensional recognition to the object.
2. Related Art
Conventionally, a method for imaging a real object with plural cameras and acquiring (restoring) the three-dimensional information of the object based on plural pieces of image data obtained by the imaging of each camera is proposed as one of an image processing technology. Processing of recognizing (specifying) a type of the object or processing of specifying a position or a posture of the object is performed using the three-dimensional information.
In such cases, three-dimensional model data in which the object is defined is previously prepared, and the three-dimensional model data is checked against the acquired three-dimensional information, thereby specifying the type, position, and posture of the object.
For example, Japanese Unexamined Patent Publication No. 2000-099760 discloses a method for producing a three-dimensional physical body model based on a stereo image that is obtained by observing a physical body from different directions. More specifically, in the method, stereo measurement is performed to a real model of a recognition target from plural directions, and the pieces of three-dimensional information acquired by the measurements are aligned and unified to produce a geometrical model of the whole physical body.
There is disclosed a technique of recognizing the position or posture of the physical body by checking the acquired three-dimensional information against the previously-registered three-dimensional model data (see Yoshihiro Kawai, et al., “Stereo Correspondence Using Segment Connectivity”, Transaction of Information Processing Society of Japan, August 1999, Vol. 40, No. 8, p. 3219-3229)
However, in the three-dimensional physical body producing method disclosed in Japanese Unexamined Patent Publication No. 2000-099760, because the three-dimensional model is produced by imaging the object that exists in reality, unfortunately there is a significant processing amount necessary to produce the three-dimensional model. In the technique called “segment-based stereo” (see Yoshihiro Kawai, et al., “Stereo Correspondence Using Segment Connectivity”, Transaction of Information Processing Society of Japan, August 1999, Vol. 40, No. 8, p. 3219-3229), an edge in the stereo image is divided into units called “segments”, and stereo correlation evaluation is performed in units of segments to acquire three-dimensional information on an outline. Therefore, when a condition to image the object varies temporally, unfortunately the three-dimensional model is produced while an error is accumulated.